The long-term goal of this research program is to reduce relapse among cocaine abusers by investigating more effective treatment interventions. The theoretical framework guiding this work is based on: 1) a coping skills model of relapse derived from social learning theory, and 2) the belief that finding key individual difference variables for matching patients to treatment will enhance treatment efficacy. Specific aims of this study are to test the effectiveness of family social skill training derived from social learning theory for treating cocain abusers, and to examine variables of possible utility in future patient-treatment matching studies. These variables include differences in coping skills, family history of substance abuse, depression, trait anxiety, and concurrent alcohol abuse. A two-group controlled design will be used, with family social skills training being compared to an attention-placebo control group (family discussions) when both are added to standard inpatient substance abuse treatment. Assessment includes multiple measure of drug and alcohol use at 6, 12, and 18 month follow-up interviews, measures of coping skills on role-play tests, and measures of family and psychosocial functioning. These assessments will allow evaluation of treatment effects and relapse, and will also permit evaluation of the role of various prognostic and mediating variables.